Meituan LongCat-2.0 fully open source under MIT license, 1.6T MoE model releases open weights and inference code
Original title:🐱 LongCat-2.0 is now fully open-source — MIT licensed, no restrictions. Since our launch a few day...
AI summary
Meituan today announced that LongCat-2.0 is fully open source (MIT license), publishing model weights and inference code. The model uses a MoE architecture with 1.6T total parameters, about 48B activated per token, and support for a 1M token context. Technical highlights include LongCat Sparse Attention for efficient long-text processing, Zero-Compute Experts that dynamically activate 33B-56B with zero wasted computation, and MOPD routing tasks to three expert groups: Agent, Reasoning and Interaction. Benchmark results: Terminal-Bench 2.1 70.8; SWE-bench Pro 59.5 (surpassing GPT-5.5's 58.6); SWE-bench Multilingual 77.3; FORTE 73.2; RWSearch 78.8; BrowseComp 79.9. It natively integrates tools such as Claude Code, OpenClaw and Hermes Agent, supports GPU and NPU deployment, and has been validated on large-scale domestic clusters.
Key Takeaways
- Meituan today announced that LongCat-2.
- 0 is fully open source (MIT license),publishing model weights and inference code.
- The model uses a MoE architecture with 1.
- 6T total parameters,about 48B activated per token,and support for a 1M token context.
- Technical highlights include LongCat Sparse Attention for efficient long-text…
Quick Facts
- Published
- Updated
- Category
- Model Releases
- AI score
- 81
国内大厂首个在 SWE-bench Pro 上超过 GPT-5.5 的开源模型,MIT 协议无任何限制,搞 agentic coding 的团队可以直接用,是个重要转折。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Meituan LongCat-2.0 fully open source under MIT license, 1.6T MoE model releases open weights and inference code. July 5, 2026. https://new.nat6.net/en/news/cmr7vpphy007oslgnz4aotpx4/